x-ray imaging - queen's...
TRANSCRIPT
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray imaging
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Würzburg, November 8, 1895: Dr. Wilhelm Conrad Röntgen
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Heat
=DC
Fluorescence in glass tube by cathode rays
--
---
--
Glowing light in gas
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Röntgen’s experiment
Heat
=DC
--
---
--
Black cardboard box
Faint glowing
Barium Platinum Cyanide
50cm
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
What did Röntgen know and did not know?
DID KNOW• Ray-like phenomenon (named it X for “unknown” ray)• Not charged• Carries to some distance• Penetrates through regular materials• Absorption rate depends on material density• Absorption depends on material mass• Carries significant energy• Energy changes with tube voltageDID NOT KNOW• Electromagnetic wave, just like visible light• Shorter wavelength (E=h*n)• Always harmful, potentially lethal
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Frau Röntgen’s hand, December, 1895
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Chest X-ray, 1903
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Army hospital 1915
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Bucky Grid (1913)
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Potter’s Grid (1920)
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Collimator grid
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray of a hand1895
2002
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
••
•
Modern chest X-ray machine
Number of X-rays registered by Medicare in 1993:
207,753,747
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray process (in its full glory)
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-Ray dermatitis & protection1910
1914
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Figure 24. An important, and adjustable, determinant of image quality in radiography is the voltage
applied across the X-ray tube by the generator. The image to the left, (a), was made at a lower voltage
(70,000 V) than (b), the one on the right (110,000V). As a result, (a) more clearly shows the two ureters,
which are carrying urine (and iodine-based contrast agent) from the kidneys to the bladder. The lower the
voltage, the better the contrast, but the greater the amount of radiation dose that is delivered to the
patient-one of many trade-offs to be considered in imaging. Courtesy of the American College of
Radiology.
Effect of tube voltage
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Decay of intensity in matter
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray interaction with DNA
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Radiation risk
• Generally, risk is the possibility (not probability) that something bad may happen
• Risk is subjective and the public’s feeling depends on the source of risk– Human made hazards are less acceptable
– Risk feels greater if comes without consent or consultation
– Accidents increase the feel of risk
– Terrorism is the supreme tool in creating extreme feeling of risk
• Radiological risk: the possibility that X-ray used in medical procedures may give rise to cancer.
• It is real, documented to happen, and objectively measurable.
• Risk = Dose * Risk-probability
• Risk-probability depends on the type of cancer & organ – how much dose over how long a time the organ tolerates before (w/ some probability) it becomes cancerous
• Yardstick: whole-body cosmic background radiation (BGR) 300 mrem /year.
• Chest X-ray: 20 mrem upon skin entry, 2 mrem upon exit
• Equivalent cancer risk:
– For Skin : Cosmic BGR = 15 X-ray shots /year
– For most internal organs: Cosmic BGR = 150 X-ray shots /year
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray = terror and horror
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Dose escalation models
• Under perennial debate
• It has been linear accumulation model: the body never forgets radiation
• But, the body does forget radiation, we just do not know how
• How does genetic distortion caused by ionizing radiation fades out? – We do not know
• “Annual limits” for medical staff
• Dose tolerance levels are set for organs
• Sensory organs are very sensitive
• Brain, lymph nodes, bone marrow are generally sensitive
• Large differences in tolerance, even within one organ (brain)
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Typical X-ray imager modelPoint source & cone beam
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray projection
X-ray source (point source)
Project an object onto the
detector to obtain an image
(shadow) of the object on the
detector.Object
Image
Detector
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray magnification
X-ray detector
X-ray source (point source)
beam central
Object
Shadow
SDD (source-detector distance)
SOD (source-target plane distance)
Linear magnification ≈
Areal magnification ≈
SDD/SOD
(SDD/SOD)2
Target plane
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Loss of 3rd dimension (depth)
X-ray detector
X-ray source (point source)
Target ?
Shadow
From the X-ray image only, we
cannot tell the location, shape and
size of the target
Because:• Projection has no inverse
• Projection is irreversible
• Projection loses sense of
depth
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Back projection (BP)
X-ray detector
X-ray source (point source)
Draw lines from the object’s
shadow back from detector
toward the X-ray source.
The object must sit on these
lines somewhere...
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Rotational X-ray & coordinate system
Source
X-ray
detector
X-ray
coord
system
Detector
coord
system
Beam
Central
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Typical mobile rotational X-ray (C-arm)
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray reconstruction of small target
• Take two or more x-ray shots in different X-ray
poses
• Keep the patient stationary!
• Find centers of target images (T1, T2., … Tn.)
• Write up the equations of all BP lines
• Compute intersection of all BP lines
S1S2
T1
T2
T
BP1BP2
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Why BP lines do not intersect?
• BP lines in space never intersect but can come close
• Some sources of error :
– Target Localization Error in detector
– Target are too large - point model is wrong
– Image warping
– Mechanical imperfections
• SSD, SAD, Detector center not known accurately
• Rotation angles not known accurately
• Patient moves between shots
• X-ray source/detector motion between shots
– Axis wobbling
– Mechanical deformation
– Accidental motion on rubber wheels
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Target correspondence problemSmall similar targets
Which belongs to which?
Brute force solution: assign them in all possible
combinations; compute reconstruction; compute
reconstruction error – the error is the smallest when the
correct correspondence.
Works for small number of targets that are quite
distantly situated. (NP hard problem, exponential
blow.) Dog shot with 100+
BB gun slugs
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Epipolar constraint to solve correspondance
S1S2
G
B
G2
B2
B1
G1
Thought experiment: replace S2-G2 line with a thin
“wire”, and use S1 source to project the wire onto D1
detector. The result is the so called epipolar line.
Since B lies somewhere on the “wire”, B1 must lie on
the projection image of the wire (epipolar line). G1 is
probably not on the epipolar line – we can use this to
positively match G1 and G2.
Epipolar line
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
When hell breaks loose…
Radioactive seed implants in the prostate
brachytherapy: Imagine over 100 rice grains in a
golf ball
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray reconstruction of large solid target• We assume the target to be a dense object
• Take many x-ray shots
• We weep the object (patient) stationary
• Reconstruct the approximate center of the
target: • Find centers of target images (T1, T2., … Tn.)
• Write up the equations of all BP lines
• Compute intersection of all BP lines
• Next, we reconstruct the shape as well as we
can…
S1S2
T1
T2
T
BP1BP2
T1
Step-1
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray reconstruction of large solid target• We put a super-cube around target’s centre
that encompasses the target
• We subdivide the cube to many tiny-tiny
cubelets (voxels)
• We forward project each voxel onto ceach
detector. If the shadow of the voxel falls
outside the target’s countour, the voxel does
not belong to the target – RED cubelet
• If in all images, the voxel’s shadow falls inside
the target, the voxel belongs to the target –
GREEN voxel
• We repeat this for each voxel…
S1S2
T1
Step-2
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray reconstruction of large solid target• Having repeated the test for all voxels in the
super-cube, we and up with a pile of green
voxels representing the inside of the target
surrounded by red voxel outside the target.
• We may connect the voxels into a solid object,
if need to.
• We can use convex hull – but that obliterates
all concavities
• From large number of images, we can
approximately reconstruct a shape.
• We cannot reconstruct dimples and holes.
• But some concave shapes like a kidney can
be reconstructed.
• We may not want to reconstruct a detail that
we cannot treat anyway.
S1S2
T1
Step-3
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray reconstruction of large solid targetOPTIMIZATION:
• Reconstruction quality and time increase
with
• Smaller voxel size
• Larger number of X-ray shots
• Reduce number of “empty” voxels by:
• Making the super-cube the smallest
• Use the tightest fitting super sphere
S1S2
T1
Step-4
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Delay between shot and image
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Real-time X-ray = fluoroscopy
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Edison with his fluoroscope, 1896
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
The first commercial fluoroscope
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray + Television = Fluoroscopy
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
X-ray image intensifier tube
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Mobile C-arm fluoroscopes
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Some fluoroscopy images
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Image warping
Courtesy of Yao & Taylor
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Flat panel detectors
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Flat panel C-arm fluoroscope
Laboratory for Percutaneous Surgery (The Perk Lab) – Copyright © Queen’s University, 9/26/2017
Summary: Pros and cons of X-ray
• Pros:
– Simple
– Inexpensive
– Excellent bone contrast
• Cons:
– Harmful dose to patient
– Loss of shape and depth of objects
– Poor soft tissue contrast